Models are being investigated that can predict survival after donor kidney grafting. The analysis of back data helps to understand the impact of many variables
Among the infinite applications of artificial intelligence in the medical field, those concerning the possibility of predicting the future outcome of a transplant with a certain degree of accuracy are among the most interesting. Now an international study, sponsored by Translational Research Centre for Organ Transplantation (Inserm, Parcc, Universit de Paris), led by professor Alexandre Loupy, and published in Lancet Digital Health (article with details here), led to develop a new system for predicting survival after a kidney transplant. The model was renamed Dispo, acronimo per Dynamic, integrative system for predicting outcome, or a dynamic and integrative system for forecasting the outcome. It is, as the authors explain, a dynamic approach of artificial intelligence to improve risk stratification for kidney transplant recipients by generating continuously refined predictions of survival through constantly updated clinical data.
Chronic kidney disease
There is a premise to make: chronic kidney disease (Mrc) a very widespread disease in the world, with an increasing prevalence in the general population. This worldwide uniform phenomenon e it is estimated that about 10% of the population of both developed and developing countries is affected by Mrc in most cases misunderstood. MRc can be caused by various causes, and in its most severe forms, based on 2020 data, affects more than 7 million people around the world. Health agencies and scientific societies have repeatedly stressed the need to develop a prediction model for survival after allogeneic kidney transplantation adapted to routine clinical practice that would improve patient decision making and management.
Over the past 20 years, various types of kidney transplant survival predictions have therefore been proposed. These predictive models are based precisely from the methodological point of view on the analysis of retrospective data: that is case series data from various transplant centers are collected and the impact of various variables is analyzed (e.g. underlying disease, patient and donor age, immunological compatibility, eventual rejection episodes, immunosuppressive therapy, and renal function over time) may affect transplant results. The novelty of this work not so much in this type of approach, as in the type of statistical model used, which it introduces not only some punctual variables (at the transplant), but also some dynamic variables, observed post-transplant, he explains Massimo Cardillo, director of the National Transplant Center (Cnt). Data from nearly 14,000 adult kidney transplant recipients from 18 academic transplant centers in Europe, the United States and South America and a cohort of patients from six randomized controlled trials were used in this observational study. Result? According to the authors, their system showed good prediction performance. We believe it could therefore help refine the prognostic assessment of kidney transplant recipients in routine clinical practice, improving precision medicine and individualized patient management. To evaluate the effect of Dispo in clinical practice, we will conduct a randomized controlled trial, they add.
A feasible approach
The group that conceived and conducted the study among the most important in the world and among the pioneers in the application of AI in transplant medicine, underlines Giuseppe Orlando, transplant surgeon and researcher at Wake Forest University, USA. The experimental model, but it definitely demonstrates the feasibility of the approach. It will be interesting to see what future developments will be, in terms of how and if they can offer it as a standard of care in France or elsewhere. I don’t see any limits at the moment, as long as the data produced by this approach are ultimately always interpreted by the transplantologist who must always have the last word, Orlando is keen to point out.
And are these methods used in Italy?
We don’t have an application which is used systematically to make predictions. We have for the information that we derive from the national data collected and published by the Cnt, which for years has activated a program for monitoring the results of transplants in all Italian centers, so we know that a certain transplant with certain characteristics has a predictable average life of a certain type, replies Cardillo. I believe that the use of data provided by this kind of applications can be useful in communication of the caregiver with the patients, but also that this must be done with great caution, because forecasting models cannot always take into account the complexity of individual cases and all the variables that affect the outcome of the transplant, he concludes.
January 31, 2022 (change January 31, 2022 | 19:50)